Install
$ agentstack add skill-jed-tech-spar-kit-act ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Implement change
Execute the plan in plan.md while strictly preserving the intent in _spec.md. Work sequentially, validate each step, keep progress updated. Refine or reorder tasks only for correctness; do not change underlying intent without approval. No silent workarounds, no drifting from the spec.
Inputs
Docs//Specs//_spec.mdDocs//Specs//plan.md- Repository code and docs
Outputs
- Code changes that satisfy
_spec.md - Updated
plan.mdwith progress - Proposed
_spec.mdupdates (for approval only—do not apply without approval)
Workflow
1. Load context
- Read
_spec.mdandplan.md - Confirm scope and change folder
- Identify current progress (checked tasks)
2. Execute tasks in order
For each unchecked task:
- Understand the task and affected files
- Implement the change
- Validate (build/run/tests or targeted checks as appropriate)
- Mark the task complete in
plan.md
Proceed sequentially unless reordering is clearly required for correctness.
3. Validation
- Prefer targeted checks when a full build is unnecessary
- Ensure changes align with Success Criteria in
_spec.md - Ask the user to test or validate only if the agent cannot run the needed checks
4. Failure handling
If a task fails:
- Fix obvious issues and retry once
- If still failing: stop, explain the blocker, propose next steps, ask how to proceed
Do not introduce workarounds that change intent or degrade quality without approval.
5. Task evolution (allowed)
- Refine wording for clarity
- Split tasks into smaller steps
- Reorder tasks if needed for correctness
Do not alter task intent in a way that would change the spec.
6. Spec change handling
If implementation reveals incorrect assumptions, missing constraints, or a better approach that changes intent:
- Propose the spec update clearly
- Do not implement that divergent path yet
- Wait for user approval
- Update
_spec.mdonly after approval
7. Keep context accurate
- Keep
plan.mdin sync with progress - Add brief notes for edge cases, follow-ups, or partial work
Stop conditions
Stop and ask if:
- A task cannot be completed after one retry
- A required spec change is identified
- The task list is insufficient or conflicts with the spec
Completion
Implementation is complete when:
- All tasks are completed or explicitly resolved
- Code satisfies Success Criteria in
_spec.md - No unresolved blockers remain
Then:
- Summarize: tasks completed, deviations or notes, validation results. Be concise.
- Ask: Should I clean up and proceed with documentation?
- If yes: begin the
document-changeskill if available; otherwise document per repo conventions
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Jed-Tech
- Source: Jed-Tech/spar-kit
- License: Apache-2.0
- Homepage: https://jed-tech.github.io/spar-kit/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.